Results: Gene level
60% of actual genes predicted completely correct
5-10% missed genes (comparable to Sanger Center)
40% wrong genes, a lot of short genes over-predicted (possibly not annotated in Standard 3)
Splitting genes is a bigger problem than joining genes